(PDB-2256) Fix terminus bug with large binary catalog data #1792
Conversation
The terminus includes code to help users diagnose the source of data that cannot be converted to UTF-8. There are several sources of this, but one is having an incorrect (or known) character set for portions of the catalog. When this invalid character data is large, it would cause the terminus to hang trying to include debugging information. This patch changes the terminus to only look for the first instance of bad data. It will also avoid doing the extra calculations needed for the error context unless debug mode is enabled. When not in debug mode there should be no impact on performance.
There was a problem hiding this comment.
This new code only checks the index of the first invalid character, just like https://github.com/puppetlabs/puppetdb/pull/1789/files#diff-a0488ae759c41269703997e51a97b388R102
|
@vzctl @olivierHa can you try this out and see if it fixes your problem? I used a manifest like the one described here and it's working fine for me. |
|
Refer to this link for build results (access rights to CI server needed): |
|
Yes, it fixes the problem, there are no performance issues with and without debug output. Thanks @senior for fixing it up! |
|
Testing this now 💃 |
|
Works for me using the catalogs for testing the original patch as well as some new ones generated per the ticket description 👍, any idea what's going on with travis @senior? |
…ta-causing-terminus-hangs (PDB-2256) Fix terminus bug with large binary catalog data
|
Travis looks to be an unrelated timeout, gonna merge this in |
5 participants
The terminus includes code to help users diagnose the source of data
that cannot be converted to UTF-8. There are several sources of this,
but one is having an incorrect (or known) character set for portions of
the catalog. When this invalid character data is large, it would cause
the terminus to hang trying to include debugging information.
This patch changes the terminus to only look for the first instance of
bad data. It will also avoid doing the extra calculations needed for the
error context unless debug mode is enabled. When not in debug mode there
should be no impact on performance.